PHYSICA MACEDONICA
-
Upload
others
-
View
15
-
Download
0
Embed Size (px)
Citation preview
INSTITUTE OF PHYSICS
Sts. Cyril and Methodius University, Skopje, Macedonia
: - (), - , - , - , - . EDITORIAL BOARD: D-r Blagoja Veljanoski
(president), d-r Nace Stojanov, d-r Nenad Novkovski, d-r Olga
Galbova, d-r Dragan Jakimovski. : - , , , . 162, 1000 , ADRESS:
Sts. Cyril and Methodius University, Faculty of Natural Sciences
and Mathematics, Institute of Physics, P.O.Box 162, 1000 Skopje,
Republic of Macedonia The journal computer preparation is done at
the Institute of Physics.
Physica Macedonica Vol. 62 pp. 1 - 47 Skopje 2016
C O N T E N T S
1. B. Radiša, M. Mitrovi and B. Misailovi
.................................................................................
1 THE EFFECT OF SOLUTION PRE-HISTORYON CHIRALITY OF SODIUM CHLORATE
CRYSTALS FROM AQUEOUS SOLUTIONS
2. S. Rendevski, I. Aliu and N. Mahmudi
......................................................................................
7 NORMALIZED INTEGRAL FOURIER TRANSFORM AMPLITUDE FOR FOOD QUALITY
DETERMINATION BY ACOUSTIC IMPULSE RESPONSE METHOD
3. D. Krstovska and B. Mitreska
.................................................................................................
13 NONLINEAR ULTRASOUND WAVE GENERATION IN QUASI-TWO DIMENSIONAL
ORGANIC CONDUCTORS
4. O. Atanackovi
........................................................................................................................
22 ON TWO EXTREMELY FAST METHODS FOR THE SOLUTION OF THE NON-LTE
LINE FORMATION PROBLEM
5. Luljeta Disha, Defrim Bulku and Antoneta Deda
...................................................................
34 METHOD OF ABSOLUTE GRAVITY MEASUREMENT AND ITS INFLUENCE IN MASS
DETERMINATION
6. N. Mahmudi and S. Rendevski
...............................................................................................
41 GAMMA RADIATION SYNTHESIS OF POLYMERS AND CONTROL OF THE
MECHANICHAL PROPERTIES
Physica Macedonica 62, (2016) p. 1-6
THE EFFECT OF SOLUTION PRE-HISTORYON CHIRALITY OF SODIUM CHLORATE
CRYSTALS FROM AQUEOUS
SOLUTIONS
University of Belgrade-Faculty of Physics, Studentski trg 12, 11000
Belgrade, Serbia
Abstract. The phenomenon of spontaneous chiral symmetry breaking
attracts the attention in physics, chemistry and biology. Sodium
chlorate crystals are an interesting example of optically active
system. In this paper results of investigations of effect of growth
history on sodium chlorate crystals chirality will be presented.
Crystals were nucleated and grown at T = (28.00±0.02)°C from
aqueous solution saturated at T0 = (31.0±0.1)°C. Prior to
nucleation the solution temperature firstly increased to Tinc, and
then slowly decreased to growth temperature. Obtained results
showed that solution pre-history induced the increase in the
population of only one of enantiomers, i.e. D-crystals dominated.
Results indicate that solution pre-history might affect chiral
symmetry breaking in crystallization of sodium chlorate crystals
even achieving crystals of chiral purity.
PACS: 81.10.Aj, 81.10.Dn
1. INTRODUCTION
Chiral symmetry breaking happen when a chemical and physical
processes who do not have any priority in production of one or the
other of the enantiomer spontaneously give rise to a large number
of one of the enantiomers: left-handed (L) or right-handed (D).
Considering the energy, there is equal probability for creating and
existence both enantiomers. Chiral asymmetry is found everywhere in
nature, and one example are living organisms who utilize L-amino
acids and D-sugars. The fact that living organisms choose only one
type of amino acids and only one type of sugars as a constituent
component has intrigued many scientists. Mechanisms which lead to
total chiral purity or to transition from one enantiomer to another
are of the great interest.
Sodium chlorate (NaClO3) molecules are achiral but they form chiral
crystals in the cubic space group P213.[1] Chiral crystals of
sodium chlorate are optically active, and L or D crystals can be
easily identified with a pair of polarizers. When polarizers are
crossed crystal color changes with rotation of the analyzer (Fig.
1a), i.e. the color of the crystal will be different for clockwise
or counterclockwise analyzer rotation depending on the handedness
of the crystal (Fig 1b).
Kondepudi et al.[2] showed that stirring during the crystallization
of NaClO3 from aqueous solution was sufficient to produce more than
99% of crystals of the same handedness, while in unstirred solution
equal number of L and D crystals was obtained. Viedma[3] found
that
© Institute of Physics, 2016
Physica Macedonica, Vol. 62
adding glass flasks and stirring the solution which is consisted of
the same number of L and D sodium chlorate crystals, leads to
complete deracemization of NaClO3, i.e. one enantiomer disappears
completely. Chiral symmetry breaking was also reported in
experiments with solution temperature changes (heating and cooling
cycles)[4] and for other substances [5,6].
Figure 1. a) Crystal color changes with analyzer rotation b) L and
D sodium chlorate crystals in solution.
These results are experimental evidence of chiral symmetry breaking
on macroscopic scale induced by autocatalysis and tendency for
domination of one enantiomer group. One of explanations of the
process of chiral symmetry breaking is secondary nucleation.
Namely, it is considered that nucleation of chiral ″mother″ crystal
influence the production of generations of secundary nucleated
crystals of the same handedness.[7-17] In the paper[18] it was
shown that chiral symmetry breaking occurs in primary nucleation,
and that secondary nucleation only increase it. Due to that,
process of chiral symmetry breaking cannot be explained by the
nucleation of ″mother″ crystal, i.e. mechanisms which induce
domination of one enantiomer group are still unclear.
In order to explain it, different models were suggested and most of
them take in consider first phases of the crystallization. Niinomi
et al.[19] showed that in crystallization process driven by
evaporation of solvent, achiral metastable crystals of sodium
chlorate formed prior to chiral crystals in solution growth, and
that transition from achiral to chiral state occurred, which means
that the chirality of the NaClO3 crystal emerges during the
transition. Differences in solubility of metastable achiral and
chiral NaClO3 crystals are also important for describing the
phenomenon. [20]
2. EXPERIMENTAL PROCEDURE
All experiments were performed with a saturated solution prepared
using the following procedure. An aqueous solution of NaClO3 was
prepared by dissolving NaClO3 (Analytical grade of substance was
99% purity) in distilled water at room temperature. Then the
solution was heated under stirring with a magnetic hot plate
stirrer and then left for a three days at saturation
2
Biljana Radiša, Mio Mitrovi, Branislava Misailovi: The effect of
solution pre-historyon chirality of sodium chlorate crystals from
aqueous solutions
temperature to precipitate the excess solute and thus to
equilibrate the solution. Crystals were nucleated in the cell by
introducing air bubbles through a needle into the cell.
We observed in situ crystal growth by digital optical microscope
Nikon SMZ800 equipped with a video recording system (camera
Luminera, Infinity 1), using a polarized light. Between 15 and 33
crystal nuclei, sufficiently distant to avoid intergrowth during
the growth, were preselected for growth rate measurements during
each growth run. In order to investigate the influence of solution
history on sodium chlorate crystals growth rate and chirality two
types of experiments were performed. In all experiments the
solution was saturated at temperature
C)1.00.31(0 °±=T , while the crystal growth was performed at
C)1.00.28(r °±=T . Both types of experiments consisted of two
parts. In the first part solution temperature
was increased to Tinc=35.0°C (the first type of experiments) or to
Tinc = 32.8°C (the second type of experiments), and after that
slowly decreased to the growth temperature C)1.00.28(r °±=T . The
temperature decreasing in the first type of experiments was about
2.5 h, while in the second type was about 4h. Procedure of both
types of experiments was similar and the only difference was in
initial and the final intervals of temperature change, showed in
Table 1. Temperature was changed from 1T to 2T in steps of T every
5 min. From C8.30 ° to C0.28 ° the solution was left to be cooled
by itself, in Type 1 experiments, and also for Type 2 experiments
in temperature range from C6.29 ° to C0.28 ° .
Table 1. Temperature changes
Type 1 experiment Type 2 experiment C][ 1 °T C][ 2 °T ]C[°T C][ 1
°T C][ 2 °T ]C[°T 0.35 0.34 5.0 8.32 2.32 3.0 0.34 8.32 4.0 2.32
6.31 2.0 8.32 2.32 3.0 6.31 4.31 1.0 2.32 6.31 2.0 4.31 6.29 05.0
6.31 4.31 1.0 - - - 4.31 8.30 05.0 - - -
Procedure of the second part of both types of experiments was the
same. After the temperature of growth was reached, crystals were
nucleated by introducing air bubbles through a needle into the
cell. Crystal growth lasted about four hours and in all experiments
crystals sizes were measured every 40 minutes.
3. RESULTS AND DISCUSSION
At different experimental conditions growth rate dispersions
occurred. Histograms representing these dispersions are presented
in Figure 2 and Figure 3. These dispersions are described by simple
normal distribution, also included in Figure 2 and 3. Zero growth
rates, whose existence is completely unclear,21 are excluded from
fitting procedure. Several high growth rates at the end of
dispersions, which probably pertain to higher activity of dominant
dislocation group, are excluded from fitting procedure,
too.22
3
Physica Macedonica, Vol. 62
Number of measured growth rates N, number of measured growth rates
of left-handed crystals NL and number of measured growth rates of
right-handed crystals ND, as well as the most probable growth rate
R1, R2, R3 and R4 for different experiments are presented in Table
2.
It can be seen from Figures 2 and 3, and Table 2 that crystals
grown in Type 2 experiments had higher growth rates. This is in
accordance with lattice strain theory.23 Namely, the temperature
changes were performed in smaller steps, so crystals had less
defects of structure, and consequently higher growth rates.
Table 2. Experimental results
Type 1
N=340 22±4 47.5±0.7 69±2 90±2 NL=155 26±4 51.3±0.5 ND=185 21±3
42.8±0.3 65±3 93±4
Type 2
N=160 25±4 51.8±0.7 83±3 NL=18 47.9±0.5
ND=142 25±1 53.0±0.4 82.8±0.8
Theoretical analysis shows that there is equal probability for
creating and existence both enantiomers. Results presented in Table
2 show that in both types of experiments left- and right- handed
crystals existed. In Type 1 experiments numbers of L and D crystals
are comparable, while in Type 2 experiments right-handed crystals
dominated. Due to the fact that in Type 2 experiments the number of
L crystals is significantly less than the number of D crystals,
conditions might lead to total chiral purity. Also, there was no
transition occurred from one enantiomer to another
enantiomer.
Regardless this, the influence of crystal growth history is
obvious. Further investigations in this field might bring clearer
results, and contribute to crystal growth theories
development.
Figure 2. Histograms representing growth rate in direction
<100> for Type 1 experiments of: a) all observed crystals b)
L crystals c) D crystals.
4
Biljana Radiša, Mio Mitrovi, Branislava Misailovi: The effect of
solution pre-historyon chirality of sodium chlorate crystals from
aqueous solutions
Figure 3. Histograms representing growth rate in direction
<100> for Type 2 experiments of: a) all observed crystals b)
L crystals c) D crystals.
4. CONCLUSIONS
Mechanisms which lead to total chiral purity are of the great
interest. In this paper results of investigations of effect of
growth history on sodium chlorate crystals chirality are presented.
Two types of experiments were performed. Obtained results show that
at different experimental conditions growth rate dispersions is
occurred. In experiments Type 2 crystals grown at higher rates. In
both types of experiments L and D crystals existed. Solution
pre-history induced the increase in the population of D-crystals in
experiments Type 2. In experiments performed there was no
transition from one enantiomer to another. Obtained results
indicate that solution pre- history might affect chiral symmetry
breaking in crystallization of sodium chlorate crystals even
achieving crystals of chiral purity, which request further
investigations.
REFERENCES
[1] W.H.Z.Zacharlasen, Kristallograhie 71, 517–529 (1929) [2] D. K.
Kondepudi, R. J. Kaufman and N. Singh, Science, 250, 975–976 (1990)
[3] C. Viedma, Phys. Rev. Lett., 94, 065504 (2005) [4] K.
Suwannasang, A. E. Flood, C. Rougeot and G. Coquerel, Cryst. Growth
Des. 13, 3498–
3504 (2013) [5] D. G. Blackmond, Proc. Natl. Acad. Sci. U. S. A.,
101, 5732–5736 (2004) [6] J. E. Hein, B. H. Cao, C. Viedma, R. M.
Kellogg and D. G.Blackmond, J. Am. Chem. Soc.,
134, 12629–12636 (2012) [7] J. M. McBride, R. L. Carter, Angew.
Chem., Int. Ed. Engl. 30, 293−295 (1991) [8] D. K. Kondepudi, K. L.
Bullock, J. A. Digits, J. K. Hall, J. M. Miller, J. Am. Chem.
Soc.
115, 10211−10216 (1993) [9] R. Y. Qian, G. D. Botsaris, Chem. Eng.
Sci. 52, 3429−3440 (1997) [10] R. Y. Qian, G. D. Botsaris, Chem.
Eng. Sci. 53, 1745−1756 (1998)
5
Physica Macedonica, Vol. 62
[11] T. Buhse, D. Durand, D. Kondepudi, J. Laudadio, S. Spilker,
Phys. Rev. Lett. 84, 4405−4408 (2000)
[12] R. Y. Qian, G. D. Botsaris, Chem. Eng. Sci. 59, 2841−2852
(2004) [13] D. K. Kondepudi, C. Sabanayagam, Chem. Phys. Lett. 217,
364−368 (1994) [14] B. Martin, A. Tharrington, X.-I. Wu, Phys. Rev.
Lett. 77, 2826−2829 (1996) [15] J. H. E. Cartwright, J. M.
García-Ruiz, O. Piro, C. I. S. Díaz, Tuval, Phys. Rev. Lett.
93,
035502 (2004) [16] J. H. E. Cartwright, O. Piro, Tuval, I. Phys.
Rev. Lett. 98, 165501 (2007) [17] M. Uwaha, J. Phys. Soc. Jpn. 73,
2601−2603 (2004) [18] C. Viedma, J. Cryst. Growth 261, 118−121
(2004) [19] H. Niinomi, T. Yamazaki, S. Harada, T. Ujihara, H.
Miura,Y. Kimura, T. Kuribayashi, M.
Uwaha, and K. Tsukamoto, Cryst. Growth Des. 13, 5188−5192 (2013)
[20] H. Niinomi, A. Horio, S. Harada, T. Ujihara, H. Miura, Y.
Kimura, K. Tsukamoto, J. Cryst.
Growth 394, 106–111 (2014) [21] B. M. Misailovi, D. A. Malivuk, A.
A. eki, M. M. Mitrovi, Cryst. Growth Des. 14,
972-978 (2014) [22] M. M. Mitrovi, A. A. eki, B. M. Misailovi, B.
Z. Radiša, Ind. Eng. Chem. Res. 53
(50), 19643-19648 (2014) [23] A. E. D. M. van der Heijden, J. P.
van der Eerden, J. Cryst. Growth 118, 14-26 (1992)
6
NORMALIZED INTEGRAL FOURIER TRANSFORM AMPLITUDE FOR FOOD QUALITY
DETERMINATION BY
ACOUSTIC IMPULSE RESPONSE METHOD
1Faculty of Engineering and Technology, Department of Mathematics
and Physics, Higher Colleges of Technology, Ras al Khaimah Men’s
Campus, United Arab Emirates
2 Faculty of Natural Sciences and Mathematics, State University of
Tetovo, bul. Ilinden, b.b., 1200 Tetovo, Republic of
Macedonia
Abstract. The research has been conducted with the goal of
determining the shelf-life of post harvested tomatoes by
implementing the non-destructive physical method of acoustic
impulse response. Normalized integral Fourier transform amplitude
has been introduced for enhancing the precision of shelf- life
determination in comparison to the use of the well-known firmness
parameter Fi. The firmness coefficient Fi has been calculated on
the basis of the dominant (fundamental) frequency of tomato
response signal f and sample mass m. The Fast-Fourier transform
algorithm has been used on the acoustical time dependent electrical
signals taken from an acoustical sensor attached to the product
under investigation, thus allowing the frequency dependent Fourier
transform to be found.
PACS: 43.40.+s, 45.10.-b, 46.40.-f, 87.19.r-
1. INTRODUCTION
For assessing the quality of agricultural products beyond the use
of chemical and biochemical methods, great practical importance has
been given recently to the physical methods, especially the
non-destructive methods that are fast, accurate and reliable [1.2].
One such method is the acoustic impulse response method, which has
been investigated by many researchers aiming at measuring the
firmness of fruits and vegetables [3-5]. The method gives solid
results that were proven by the classical destructive firm testing
technique [6]. The main advantage of the acoustic impulse response
method is that it is fast, reliable, non-destructive and can be
performed in-field (on-site). The main disadvantage of the method
is that it gives a “bulk” information for the product, which means
the acoustic response is like averaged for the whole sample. Only
the magnetic resonance imaging method (MRI) can give perfect
picture of the inside structure of the food product examined, but
this method requires expensive equipment and highly trained
examiner [7]. The major achievement of the acoustic impulse
response method is seen in the context of good correlation of the
results for maturity stage of food samples with the results
from
© Institute of Physics, 2016
Physica Macedonica, Vol. 62
many other methods [8]. That means that the acoustic impulse
response method can be used by alone.
2. MATERIALS AND METHODS
The “Melody” variety tomatoes (Lycopersicum) were hand-harvested
from the greenhouse orangery of the company “Agroproduct” from
Gostivar, Republic of Macedonia, on October 17, 2013. Twenty seven
tomatoes were sorted according to their mass (250 g ± 50 g) and
color appearance of same class (90% of the surface to be red and
10% of the surface showing change from green to yellow). For that
purpose, a color sensor panel have been composed of five assessors
experienced in tomato assessment. Tomatoes were harvested from the
tomato trees between 40 – 45 cm above the ground. Only one tomato
was taken from a tomato tree. The tomatoes were grown according to
the following conditions: average temperature of 14 – 16 0C during
nights; water supply of 1.5 L for 24 hours; fertilizer “Jara” of
180 g per tomato tree; pesticides treatment every week;
insecticides treatment on occasions. Tomatoes were stored at 15 ± 1
0C, relative humidity 60% and were placed in a box that can hold
each tomato separated. After been harvested, tomatoes were
transferred to the lab and kept during the investigation on average
temperature of 15 ± 1 0C and relative humidity 60 % on shelves with
artificial illumination same as under selling condition on the
local market.
According to the acoustic impulse response method, tomatoes were
impacted without a damage with a small ball pendulum and tomatoes
vibration response signals were recorded with a microphone as a
sensor. The firmness coefficient Fi has been calculated on the
basis of the dominant (fundamental) frequency of tomato response
signal f and sample mass m. The firmness coefficient Fi, which for
spherical-like fruits was first proposed by Abbott [9] and was
modified by Cooke and Rand [10], has been calculated according to
the following formula:
2 2 3
iF f m= . (1)
Same method has been applied from many researchers on various
products (apples, watermelons, mangos, and many others) [11, 12].
To reduce errors due to sample internal homogeneity differences as
well as injuries after the collision with the impactor (pendulum
ball), it is preferred to change the position of impact and the
average response is obtained after three hits on the equatorial
surface of the samples when the microphone is positioned at angle
1200 in respect to the impact [10].
The dominant or fundamental frequency of vibration of the food
product has been determined by doing Fourier transform of the
amplitude-time function of the sensor signal which gave the
amplitude-frequency function (so called power spectrum). There are
several methods described in the literature for calculating the
Fourier transform and we used the Fast Fourier Transform algorithm
incorporated in the powerful OriginPro® graphical analysis
software. This software combines some good features from MatLab®
and Mathematica®.
8
S. Rendevski1, I. Aliu and N. Mahmudi: Normalized integral fourier
transform amplitude for food quality determination by acoustic
impulse response method
Figure 1. Typical acoustic signal. Figure 2. Fourier transform of
the signal from fig.1
On figure 1, a common acoustical sensor signal is given after
impact on a tomato, and on figure 2, the Fourier transform is given
for the same function. It is obvious that the dominant frequency f
of the sample vibration has a value of 95 Hz, but there is also
another small peak in the function at 180 Hz and on some cases
there were more. The position of the main, or the dominant
frequency, and the presence of other peaks in the Fourier transform
can give more information about the food product. For example, if
there is second smaller peak, that means the vibration of the
sample is more complex, probably because of different internal
structure inside the product – already rotten regions, over-ripen
small regions that will spread in the following days, large inside
inhomogeneity in density of the product because of changes in grown
conditions, and so on. When focused only of the main peak or the
dominant frequency of vibration, the shift over time to lower
values indicates softer product and/or separations inside of the
product – one larger region with more water content and another (a
core) denser (less water content) which is typical for over-mature
stage of fruits and vegetables.
Some detailed information from the acoustic impulse response method
as well as the use of piezoelectric or microphone as a sensor
applied for tomato quality assessment after storage has been given
in literature [2-5]. In our research, we used equipment
schematically described on figure 3. Here, we have a ring sample
holder that gently hold the tomato sample; impactor with a pendulum
ball (self-made with aluminum ball of 20 g mass and pendulum angle
300 that catch the ball back after the first impact); microphone
model MI-10, A4 Tech, Taiwan, with integrated amplifier and a
computer for data acquistion and calculations. GoldWave® audio
studio software has been used for taking the amplitude-time signal
from the microphone. The sampling rate was 20,000 points per
second.
In this paper, a new quantity determined from the Fourier transform
of the acoustical signal, named normalized integral amplitude, Nia,
has been introduced (in units of Hz). The normalized integral
amplitude is obtained after integration of the Fourior transform
function on the whole frequency range and by division after by the
amplitude at the dominant frequency.
9
Physica Macedonica, Vol. 62
Figure 3. Scheme of the equipment for acoustic impulse response
measurements.
From the physics of sound and mathemaics of the Fourier transform
it is known that according to the Parseval’s theorem the energy of
the acoustical signal received by the microphone
2
( )f t dt+∞
−∞∫ is equal to the energy of the vibrational spectrum of the
sample obtained from the Fourier transform
2
( )F dω ω+∞
−∞∫ [13, 14]. Knowing this, it is obvious that Nia is directly
proportional to the energy of the signal (or the vibrational
spectrum) and after normalizing it to the maximum amplitude, Nia
can be considered as a parameter that gives information on the
“dissipation” of the energy not only around the dominant frequency
of vibration, but on the whole spectrum, indeed.
In the literature, when acusto-vibrational properties are expressed
with only the dominant frequency, the contribution to all other
frequencies of vibration in the obtained Fourier transform function
is taken as negligible. The aim here is not to connect Nia quantity
to the firmness properties of samples, as Abbott (1968) and Cooke
and Rand (1973) did, but to check the behavior of the quantity over
storage time passed.
Figure 4. Graphical presentation of the changes of f, Nia and Fi
over time.
10
S. Rendevski1, I. Aliu and N. Mahmudi: Normalized integral fourier
transform amplitude for food quality determination by acoustic
impulse response method
It is expectable to have drastic change of Nia value if several
more vibration characteristics exist (peaks) beside the dominant
one in the sample. Aiming to understand the importance of Nia
quantity, the Fourier transform function for over-matured sample
will be discussed in comparison to same sample for which the
Fourier transform function has been taken immediately after
harvesting. The values of Nia parameters were 168 Hz and 202 Hz,
respectively, while the values of the dominant frequency where 73
Hz and 102 Hz, respectively. In the research presented it has been
found that when the difference between the Nia value and f is more
than double, the product can be considered as over-matured, as in
the above example when we found f = 73 Hz and Nia = 168 Hz. When
time dependences of Nia and f is constructed, there is more drastic
change of Nia value than f value after some critical moment
(fig.4). This moment is direct indication of the shelf-life of the
tomatoes because after this critical time the quality of the sample
deteriorates significantly.
3. RESULTS, DISSCUSIONS AND CONCLUSIONS
The mass of the tomatoes are important quantities for determination
of the firmness parameter, according to Eq.1. The mass was measured
each experimental day using the electronic weighing scale with
precision of 1 g. Mass loss was calculated according to the
equation
[( ) / ] 100%l i f im m m m= − ⋅ , where ml is the reduced mass in
percentage, mi is the initial mass and mf is the final mass of the
product, in grams. It has been found during the research that the
reduced mass was about 5% of the initial mass of all tomatoes. This
is due to the water loss during the 11 days period of
investigation. However, such a loss of mass was not considered
influential on the change in maturity stage of the product in
investigated period.
The values of the dominant (fundamental) frequency of vibration f,
normalized integral amplitude Nia, firmness parameter Fi calculated
from Eq.1, are given in Table 1.
Table 1. Main parameters obtained from the acoustic impulse
response method.
Time (in days) f,Hz Nia,Hz Fi, kg2/3/s2
0 108 ± 11 239 ± 18 4017 4 103 ± 10 210 ± 20 3592 7 84 ± 15 205 ±
22 2400 9 73 ± 17 138 ± 20 1568 11 66 ± 12 142 ± 23 1458
From fig.4, it is obvious that the change of Nia over time are more
pronounced in comparison to the changes of f. The change of Nia is
following an inverse S graphical dependence from which it is
visible that after 9 days of storage there is significant decrease
of Nia value which indicates over-mature stage of the tomatoes. The
same behavior has been found when changes of the firmness parameter
Fi have been followed over days passed (fig.4). This indicates that
Nia quantity can be considered as a measure for the vibration
characteristics of the tomato, instead of the single dominant
frequency of vibration. The dependence of Nia over time is
following the dependence of Fi in more profound way than the f.
That implies that there is direct connection between Nia and Fi
that requires theoretical investigation henceforth.
11
Physica Macedonica, Vol. 62
As seen from the results from the application of the acoustic
impulse response methods, for the “Melody” type tomatoes kept at 15
0C at 60 % relative humidity and normal illumination, the
shelf-life has been found for nine days of storage.
The method of acoustic impulse response provides a “view” of the
internal homogeneity of tomatoes without destructing them and for
this reason it is considered as a method of non- destructive
testing of products. If there is inhomogeneity inside tomatoes,
that will be reflected through different dominant frequencies and
normalized integral amplitude of the product vibrations during
slight impact of the tomatoes in different positions, for example,
around the stem, at the bottom of the fruit or in the equatorial
part.
REFERENCES
[1] Gunasekaran, S. “Nondestructive food evaluation: Techniques to
analyze properties and quality”. New York, USA: CRC Press
(2000)
[2] Kilcast, D. “Instrumental assessment of food sensory quality: A
practical guide”. Cambridge, UK: Woodhead Publishing (2013)
[3] Galili, N., Shmulevich, I., Benichou, N. Transactions of the
American Society of Agricultural Engineering, 41, 399-407
(1998)
[4] Duprat, F., Grotte, M., Pietri, E., Loonis, D., Studman, C.J.
Journal of Agricultural Engineering Research, 66(4), 251-259
(1997)
[5] Aboonajmi, M., Jahangiri, M., Hassan-Beygi, S.R. Journal of
Food Processing and Preservation, 39(6), 3175-3188 (2015)
[6] Barrett, D.M., Beaulieu, J.C., Shewfelt, R. Critical Reviews in
Food Science and Nutrition, 50(5), 369-389 (2010)
[7] Zhang, L., McCarthy, M.J. Postharvest Biology and Technology,
67, 37-43 (2012) [8] Arana, I. “Physical properties of food”. Boca
Raton, USA: Taylor & Francis Group (2012) [9] Abbott, J.A.,
Bachan, A., Childers, G.S., Fizterald, J. V., Matusik, F. J. Food
Technology,
22, 635-645 (1968) [10] Cooke, J. R., Rand, R. H. Journal of
Agricultural Engineering Research, 18, 141-157
(1973) [11] Abbott, J.A. Quality of Fresh and Processed Fruits,
542, 265-279 (2004) [12] Butz, P., Hofmann, C., Tauscher, B.
Journal of Food Science, 70(9), 131-141 (2005) [13] Pain, J.H. “The
physics of vibration and waves”. Chichester, UK: John Wiley and
Sons
(2005) [14] Urumov, V. “Mathematical physics: methods, applications
and problems (in Macedonian)”.
Skopje, Republic of Macedonia: Prosvetno Delo (1997)
12
NONLINEAR ULTRASOUND WAVE GENERATION IN QUASI-TWO DIMENSIONAL
ORGANIC CONDUCTORS
D. Krstovska and B. Mitreska Institute of Physics, Faculty of
Natural Sciences and Mathematics, Ss. Cyril and
Methodius University, P. O. Box 162, 1000 Skopje, Macedonia
Abstract. The amplitude of an ultrasound wave with double frequency
2ω generated by Joule heating as a heat source in the quasi-two
dimensional layered organic conductor β−(BEDT-TTF)2IBr2 is analyzed
as a function of the magnetic field strength and its orientation
with respect to the plane of the layers. Angular oscillations of
the ultrasound wave amplitude are correlated with the changes of
electrical and thermal conductivity of the compound when the
magnetic field is tilted from the normal to the layers plane due to
the periodic dependence of the carriers velocity along the less
conducting axis on the field orientation. The period of
oscillations allows the characteristic parameters of the Fermi
surface to be determined. The magnetic field dependence reveals
that the effect of nonlinear generation of the ultrasound wave due
to Joule heating is quadratic in proportion to the magnetic field
at each field direction. The behavior of the wave amplitude is
different depending on the field orientation from the normal to the
layers, i.e., is it oriented at the maximum or minimum in the
angular dependence. In layered organic conductors due to the small
electron mean free-path the nonlinear generation of ultrasound
waves can be observed for a wide range of magnetic fields providing
possibilities for experimental studies using non-contact ultrasonic
techniques.
PACS: 74.70.Kn, 72.15.Jf, 72.50.+b
1. INTRODUCTION
The mechanisms of linear transformation, which are responsible for
the generation of ultrasound in conducting media at the frequency
of an electromagnetic wave ω incident on the surface of the metal,
are a standard subject of investigation in electromagnetic-acoustic
conversion problems [1-6]. In conducting media, the mechanism
occurs as follows: when an electromagnetic wave with frequency ω is
incident on the conductor, nonuniform temperature oscillations of
the same frequency appear as a result of the thermoelectric effect.
These oscillations, in turn, generate acoustic oscillations in the
conductor with a frequency that coincides with the frequency of the
incident electromagnetic wave (contactless excitation of
ultrasound).
The induction [7-10] and deformation force [11-18] have been
studied in greatest detail as sources of linear transformation of
waves. In the last few years the focus of investigations of
electromagnetic excitation of ultrasound has shifted to the study
of subtle nonlinear conversion mechanisms. Some studies in ordinary
metals have shown that these mechanisms could also generate
ultrasound wave with frequency 2ω, i.e., in the nonlinear regime
[19-23].
© Institute of Physics, 2016
Physica Macedonica, Vol. 62
Apart from the induction and deformation forces, longitudinal
ultrasound at double frequency can also be generated by other
source of nonlinearity. The most important such source is the force
arising due to the appearance of thermoelastic stresses in the
Joule-heated skin layer [24]. Joule heating is the characteristic
of material that can be heated up when there is current flow
because electrical energy is transformed into thermal energy. The
thermoelectric mechanism of ultrasound generation is in principle a
nonlinear interaction due to the action of the thermoelastic
stresses on the crystal lattice. These stresses that are caused by
the temperature oscillations with frequency 2ω (Θ∼cos(2ωt)) arising
from the alternating part of the Joule heat generate ultrasonic
waves with double frequency.
In this paper we study the effect of Joule heating from electric
currents flowing through a quasi-two dimensional organic conductor
on the temperature distribution and nonlinear generation of a
longitudinal ultrasound within the conductor. The quasi-two
dimensional (q2D) charge-transfer salts form the largest organic
superconductor family with unusual normal and superconducting
properties which come out of the layered structure as well as the
specific shape of the Fermi surface. Their most remarkable feature
is the reduced dimensionality of the electronic band structure
caused by the specific character of their crystal structure.
Specifically, the q2D organic conductor β−(BEDT-TTF)2IBr2 is
considered and the amplitude of the generated longitudinal
ultrasound is analyzed as a function of the magnetic field strength
and its orientation from the less conducting axis to the plane of
the layers.
2. FORMULATION OF THE PROBLEM
The ultrasound with double frequency is generated as a result of
temperature oscillations which are induced by Joule heating as a
heat source. Here we consider a case when electric currents flowing
through the conductor are along the most conducting axis (x-axis),
j = (jx, 0, 0) and the only nonzero component of the electric field
is the x−component, E = (Ex, 0, 0). An ultrasound of double
frequency is generated and propagating along the less conducting
axis (z- axis), k = (0, 0, k), and therefore all of the quantities
depend only on the z−component. The conductor is placed in magnetic
field oriented at an angle θ from the normal to layers plane, in
the yz plane, B = (0, Bsinθ, Bcosθ). A temperature oscillating with
frequency 2ω can occur if 2ωτ<<1 where τ is the relaxation
time of the conduction electrons. We study the case of normal skin
effect when the condition ,1<<lkT where l is the electron
mean-free path length and kT is the thermal wave number, is
satisfied.
The complete system of partial differential equations, describing
the generation of longitudinal ultrasound at frequency 2ω, includes
Maxwell's equations for the magnetic B and electric E field, the
kinetic equation for the nonequilibrium correction χ to the
electron distribution function, and the equations of heat
conduction and elasticity:
[ ]0
z z z ω
14
D. Krstovska and B. Mitreska: Nonlinear ultrasound wave generation
in quasi-two dimensional organic conductor
Here v and p are the electron velocity and momentum, e is the
electron charge, Θ is the high-frequency addition to the mean
temperature T of the crystal, QJ is the power density of the heat
source, C is the volumetric heat capacity, zzκ is the thermal
conductivity tensor, ks = 2ω/s is the wave vector of the ultrasonic
wave at frequency 2ω, s is the ultrasound wave velocity and β is
the volumetric expansion coefficient.
The above system of equations must be supplemented with the
corresponding boundary conditions for the heat flux and the
ultrasound amplitude at the conductor’s surface. In the case when
the heat source is the time-dependent part of the Joule heating the
boundary conditions are:
2 0 0 0| 0, | | .zz z z z
U z z z
∂ ∂ ∂ (2)
By using Maxwell’s equations one obtains the following expression
for the Joule heating
2
2 ( )
0
µ −= = =jE , (3)
where (1 )k i δ= + and 02 ( ).xxδ ωµ σ= Substituting eq. (3) into
the heat conduction equation (third equation in (1)) we obtain
the
following expression for the temperature distribution
22
zz
∂ (4)
where (1 )T Tk i δ= + and .T zz Cδ κ ω= Here Tδ is the penetration
depth of the thermal field under the conditions of a normal skin
effect.
The solution of eq. (4) is given as
2
− Θ =
− (5)
The complex amplitude of the generated longitudinal ultrasound wave
with double frequency can be written in the following form
2 2
= +
(6)
The components of the conductivity tensor which relate the current
density to the electric field can be calculated by using the
Boltzmann transport equation for the charge carrier distribution
function, based on the tight binding approximation band structure
within the single relaxation time approximation τ (second equation
in (1)) [25]. The components of the electrical conductivity tensor
are determined as follows
15
∂ =
∂∫ ∫ ∫
(7)
Here is the Planck’s constant divided by 2π, pB =pysinθ +pzcosθ
=const is the momentum projection in the magnetic field direction,
and t is the time motion of the conduction electrons in the
magnetic field under the influence of the Lorentz force dp/dt =e
(v×B).
For the quasi-two dimensional organic conductor whose electron
energy spectrum has the form
2 2
ε η +
, (8)
where m* is the electron cyclotron effective mass, η is the
quasi-two dimensionality parameter, vF is the characteristic Fermi
velocity of the electrons along the layers, c is the lattice
constant, the expressions for zzσ and xxσ take the following
form
2
tan cos ( ) and ( tan ).
θ = = +
(9)
Here 0σ coincides in order of magnitude with the electrical
conductivity along the layers in the absence of a magnetic field,
Dp = 2pF is the averaged diameter of the Fermi surface along the py
axis and h = e/m∗ τ.
The thermal conductivity zzκ is obtained by using the
Wiedemann-Franz law
2 2
3. RESULTS AND DISCUSSION
The amplitude of the generated ultrasound wave with double
frequency due to Joule heating is a function of the frequency of
applied electric current ω, the magnetic field B, the angle between
the normal to the layers and the magnetic field θ, as well as of
the characteristics of the q2D conductor (electrical and thermal
conductivity). For the q2D organic conductor β−(BEDT- TTF)2IBr2,
the interlayer transfer integral and the cyclotron mass extracted
from data on magnetic quantum oscillations are tc = 0.35 meV and m*
= 4.2me, respectively and the relaxation time is of the order of 10
ps [26]. The Fermi surface is a weakly warped cylinder and the
quasi-two dimensionality parameter is of order η = 10−3.
Figure 1 shows the angular dependence of the amplitude U2ω(θ) for
several field values with pronounced peaks at certain angles whose
height decreases with increasing angle between the normal and the
plane of the layers. There is a deep minimum for the orientation of
the magnetic field in the plane of the layers. The angular
oscillations are characteristic of the kinetic and
16
D. Krstovska and B. Mitreska: Nonlinear ultrasound wave generation
in quasi-two dimensional organic conductor
thermoelectric coefficients of layered organic conductors. The
oscillatory dependence of the ultrasound wave amplitude U2ω on the
angle θ is determined by the angular oscillations of the
magnetoresistance 1−= zzzz σρ as well as thermal conductivity zzκ
which are associated with the charge carriers motion on the
cylindrical Fermi surface in a tilted magnetic field.
Figure. 1: Angular dependence of the ultrasound wave amplitude U2ω
for T = 30 K, η = 10−3 and several field values B = 3, 4.5, 6, 8,
10 and 12 T from bottom to top.
In the vicinity of angles where the amplitude U2ω has maximum
value, ,max nn θθ = the
average drift velocity zv of charge carriers along the ultrasound
wave vector coincides with the velocity s of the ultrasound wave,
and their interaction with the wave is most effective. As a result,
at these angles the amplitude is the largest (especially this trend
is apparent for θ > 45) and U2ω
(θ) exhibits peaks in the angular dependence. The magnitude of the
peaks is modulated with changing the tilt angle θ due to the
oscillating zeroth order Bessel function )./tan(0 θpcDJ
If the vectors ks and B are not orthogonal, the average velocity of
charge carriers in the direction of propagation of the wave differs
from zero for any shape of the Fermi surface, i.e., charge carriers
drift in the direction of wave propagation. The existence of points
at which the interaction with the wave is most effective on
different trajectories leads to a resonant dependence of the
acoustic wave amplitude on magnetic field B and the angle between
the normal to the layers and magnetic field θ. At θ < π/2 the
charge carriers which are in phase with the acoustic wave
propagating along the z−axis are the carriers with time-averaged
drift velocity Fz vv η= in the direction of propagation of the
ultrasound wave.
The interaction of these electrons with the ultrasound wave is most
effective when 1<<Tsk δ under the condition of normal skin
effect, ,1<<lkT and when the q2D closed orbits are
not strongly warped. For θ < π/2 all orbits of electrons whose
states belong to the corrugated cylinder are closed and do not
contain self-intersections. Cross-sections of the corrugated
cylinder with the plane pB = const become more elongated with
tilting the field away from the z−axis. For θ close to π/2 there is
a significant transformation of the electron trajectories in the
momentum space. Due to the pronounced elongation of the closed
orbits the amplitude of the ultrasound wave U2ω rapidly decreases
with increasing angle as seen in Fig. 1.
17
Physica Macedonica, Vol. 62
For the amplitude U2ω plotted versus tanθ (Fig. 2), the same trends
are apparent, but important feature to emphasize is that the
amplitude U2ω is nonzero in the whole range of angles in comparison
to the amplitude of the fundamental wave Uω generated by the
thermoelectric forces in the linear electromagnetic-acoustic
transformation processes in q2D organic conductors [27]. This is
because of the sinθ factor in the denominator of the expression for
the amplitude U2ω (eq. 6).
Figure. 2: tan θ dependence of the ultrasound wave amplitude U2ω
for T = 30 K, η = 10−3 and several field values B = 3, 6, 8, 10,
12, 15 and 20 T from bottom to top. The inset shows the tan θ
dependence of the amplitude U2ω for T = 30 K, η = 10−3 and B = 20 T
with the corresponding Yamaji angles where the amplitude U2ω has
maximum or minimum value.
When /tanθpcD equals a zero of the zeroth-order Bessel function,
then at that angle the electrical conductivity along the less
conducting axis, ),(θσ zz will be negligible and the
magneto-resistance )(θρ zz will be a maximum. These angles are
known as Yamaji angles [28, 29]. If 1/tan >>θpcD , then the
zeros occurs at angles ,max
nn θθ = given by /tan max npcD θ =
π(n −1/ 4), n = 0, 1, 2, 3, … For )(θρ zz to be a minimum it should
be ,min nn θθ = where
/tan min npcD θ = π(n +1/ 4). According to eqs. (6) and (9), the
amplitude U2ω has a peak or dip at
angles where )(θρ zz is maximum or minimum as expected in case when
there exist only closed Fermi surface sections. This is plotted in
the inset of Fig. 2 for B = 20 T where the corresponding angles are
marked at which the amplitude U2ω has a peak or dip, respectively.
Also, the tan θ position of the peaks and dips is field
independent, i.e., does not change with increasing field. The
narrow peaks that appear in the U2ω (tan θ) dependence repeat with
a period ./2)(tan pcDπθ = The experimental measurements of the
period will allow the diameter Dp of the Fermi surface to be
determined. In addition, from the shift between the tan θ positions
where U2ω (tan θ) is minimum and tan θ positions where U2ω (tan θ)
has a peak the value of the interlayer transfer integrals tc in the
energy dispersion could be calculated.
Next we look at the curves of the amplitude U2ω as a function of
magnetic field at fixed angles. Particularly, the angles where U2ω
(θ) has maximum and minimum value, determined from the inset of
Fig. 2, will be considered. Figure 3 shows the traces of U2ω (B)
for several field
18
D. Krstovska and B. Mitreska: Nonlinear ultrasound wave generation
in quasi-two dimensional organic conductor
directions from the normal to the layers: θ = 3, 10, 15, 28, 50.05,
64.78, 75, 83 and 90. The effect of generation of the ultrasound
wave due to Joule heating is quadratic in proportion to the
magnetic field at each field direction. It is important to note
that the field behavior of the ultrasound wave amplitude U2ω is
quite distinct for max
nn θθ = and .min nn θθ = While for angles
where the amplitude U2ω has a peak in the angular dependence, for
example at the firs peak 05.50max
1 =θ in Fig. 3, it is increasing with increasing field without any
changes, for angles where the amplitude U2ω has a dip, as for
,78.64min
1 =θ there is a narrow peak in the magnetic field
dependence at lower fields (inset of Fig. 3). This peak is also
present for small angles from the z- axis but with smaller
amplitude compared to that for min
nn θθ = as shown in the inset of Fig. 3.
Figure. 3: Magnetic field dependence of the ultrasound wave
amplitude U2ω for T = 30 K and several field directions from the
normal to the layers: θ = 3, 10, 15, 28, 50.05, 64.78, 75, 83 and
90. The inset shows the low field behavior of U2ω for T = 20 K, η =
10−3 and θ = 3, 10, 15, 28, 64.78 and 90.
This implies that at angles where the amplitude U2ω is small, below
45=θ (tan θ < 1) and at ,min
nn θθ = compared to its magnitude at the peaks there is always a
narrow peak in the field dependence. The appearance of the peak
could be associated with a resonant condition in the dependence of
the ultrasound wave amplitude on magnetic field B due to the large
local temperature gradient induced by the Joule heating at lower
fields. The position of the peak and the fulfillment of the
resonant condition are field and angle dependent.
The observation of nonlinear ultrasound wave generation due to
Joule heating as a heat source is constraint by the condition
2ωτ<<1. In the organic conductor β−(BEDT-TTF)2IBr2 this
condition is always fulfilled, even at high frequencies ω = 108-109
Hz, as the relaxation time τ is very small (for β−(BEDT-TTF)2IBr2,
τ = 10 ps [26]). In addition, for the generation of waves to be the
most effective other conditions must be satisfied. This includes
fulfillment of 1<<Tsk δ as long as the conditions for normal
skin effect Tl δδ ,<< are satisfied. The Fermi velocity of
β−(BEDT-TTF)2IBr2 is vF = 1.5⋅105 m/s [30] and the electron mean
free-path is of order l = vF τ = 1.5 µm. The skin depth of the
β-(BEDT-TTF)2IBr2 crystals is about δ = 0.12 mm [30]. It is evident
that the necessary conditions are fulfilled even at higher fields
providing the ultrasound wave generation with double frequency to
be studied experimentally in a wide range of magnetic fields.
Experiments on thermoelectric generation of ultrasound waves in
layered organic conductors
19
Physica Macedonica, Vol. 62
would be highly favorable since these compounds are synthesized
with high purity that makes them very convenient for performing
experiments.
4. CONCLUSIONS
The nonlinear generation of an ultrasound wave of double frequency
(ω = 108-109 Hz) in layered quasi-two dimensional organic
conductors as a result of Joule heating as a heat source is
considered. The amplitude of the generated wave due to
thermoelectric stresses caused by the alternating part of the Joule
heat is analyzed as a function of the magnetic field B, the angle
between the normal to the layers and the magnetic field θ as well
as of the characteristics of the conductor. Specifically, the
parameters for the organic conductor β−(BEDT-TTF)2IBr2 are used to
obtain the ultrasound wave amplitude U2ω at temperature T = 30 K
and the quasi-two dimensionality parameter η = 10−3, i.e., for not
strongly warped Fermi surface. It has been shown that the necessary
conditions for observing the nonlinear generation of ultrasound
wave with double frequency are always fulfilled even in relatively
high but nonquantizing magnetic fields.
The oscillatory dependence of the ultrasound wave amplitude U2ω on
the angle θ is determined by the angular oscillations of the
magnetoresistance )(θρ zz which are associated with the charge
carriers motion on the cylindrical Fermi surface in a tilted
magnetic field. At angles where the amplitude U2ω has maximum
value, the average drift velocity zv of charge carriers along the
ultrasound wave vector coincides with the velocity s of the
ultrasound wave and their interaction with the wave is most
effective. Therefore narrow peaks appear in the angular dependence
of the amplitude U2ω that correspond to the most effective
interaction of the charge carriers with the wave. From the period
of the angular oscillations the diameter of the Fermi surface as
well as the interlayer transfer integrals in the energy dispersion
can be determined.
The magnetic field dependence of the amplitude U2ω reveals
existence of peaks present for small angles from the z-axis as well
as for angles where the amplitude U2ω has a dip in the angular
dependence. The appearance of the peak in the field dependence
could be associated with the resonant dependence of the ultrasound
wave amplitude on magnetic field B at lower fields.
REFERENCES
[1] W. D. Wallace, Int. J. NDT 2, 309 (1971) [2] E. R. Dobbs, p.
127-150, in “Physical Acoustics. Principles and Methods”, Edited by
W. P.
Mason, Volume 10, Academic, New York (1973) [3] H. M. Frost, pp.
179-275, in “Physical Acoustics”, Edited by W. P. Mason & R.N.
Thurston,
Volume 14, Academic Press, New York (1979) [4] G. A. Budenkov and
S. Yu. Gurevich, Defektroskopiya 5, 5 (1981) [5] A. N. Vasil'ev and
Yu. P. Gaidukov, Usp. Fiz. Nauk 141,431 (1983) [Sov. Phys. Usp. 26
(1
l), 952 (1983)] [6] E. Kartheuser, R. L. Ram Mohan and S.
Rodriguez, Adv. Phys. 35,423 (1986) [7] V. M. Kontorovich and A. M.
Glushyuk, Zh. Eksp. Teor. Fiz. 41, 1195 (1961) [Sov. Phys.
JETP 14(4), 852 (1962)]
20
D. Krstovska and B. Mitreska: Nonlinear ultrasound wave generation
in quasi-two dimensional organic conductor
[8] V. Ya. Kravchenko, Zh. Eksp. Teor. Fiz. 54, 1494 (1968) [Sov.
Phys. JETP 27(5), 801 (1968)]
[9] V. L. Fal'ko, Zh. Eksp. Teor. Fiz. 85, 300 (1983) [Sov. Phys.
JETP 58(1), 175 (1983)] [10] A. N. Vasil'ev, Yu. P. Gaidukov, M. I.
Kaganov et al., Fiz. Nizk. Temp. 15, 160 ( 1989)
[Sov. J. Low Temp. Phys. 15, 91 (1989)] [11] G. Turner, R. L.
Thomas and D. Hsu, Phys. Rev. B 3, 3097 (1971) [12] W. D. Wallace,
M. R. Gaerttner and B. W. Maxfield, Phys. Rev. Lett. 27, 995 (1971)
[13] D. E. Chimenti, C. A. Kukkonen and B. W. Maxfield, Phys. Rev.
B 10, 3228 (1974) [14] D. E. Chimenti, Phys. Rev. B 13, 4245 (1976)
[15] N. C. Banic and A. W. Overhauser, Phys. Rev. B 16, 3379 (1977)
[16] E. A. Kaner, V. L. Fal'ko and L. P. Sal'nikova, Fiz. Nizk.
Temp. 12, 831 (1986) [Sov. J.
Low Temp. Phys. 12, 471 (1986)] [17] E. Kartheuser and S.
Rodriguez, Phys. Rev. B 33, 772 (1986) [18] M. A. Gulvanskii, Fiz.
Tverd. Tela 29, 1249 ( 1987) [Sov. Phys. Solid State 29, 715
(1987)] [19] Yu. I. Sazonov and Yu. M. Shkarlet, Defektoskopiya,
No. 5, 1 (1969) [20] I. Cap, Acta Phys. Slov. 32, 77 (1982) [21] A.
P. Korolyuk and V. I. Khizhnyi, Pis'ma Zh. Eksp. Teor. Fiz. 48, 348
(1988) [JETP Lett.
48, 385 (1988)] [22] A. N. Vasil'ev, M. A. Gulyanskii and M. I.
Kaganov, Zh. Eksp. Teor. Fiz. 91, 202 (1986)
[Sov. Phys. JETP 64(1), 117 (1986)] [23] N. M. Makarov, F. P.
Rodriguez and V. A. Yampol'skii, Zh. Eksp. Teor. Fiz. 95, 205
(1988) [Sov. Phys. JETP 67(9), 1943 (1988)] [24] L. D. Favro, IEEE
Ultrason. Symp. Proc. 1, 399 (1986) [25] A. A. Abrikosov, pp.
201-202, in “Fundamentals of the Theory of Metals”, Volume 2,
North-Holland, Amsterdam (1988) [26] M. V. Kartsovnik, Chem. Rev.
104, 5737 (2004) [27] D. Krstovska, O. Galbova and T. Sandev, EPL
81 37006 (2008) [28] K. Yamaji, J. Phys. Soc. Jpn. 58 1520 (1989)
[29] R. Yagi, Y. Iye, T. Osada and S. Kagoshima, J. Phys. Soc. Jpn.
59 3069 (1990) [30] A. Chernenkaya, A. Dmitriev, M. Kirman, O.
Koplak and R. Morgunov, Solid State
Phenomena 190, 615 (2012)
Physica Macedonica 62, (2016) p. 22-33
ON TWO EXTREMELY FAST METHODS FOR THE SOLUTION OF THE NON-LTE LINE
FORMATION PROBLEM
O. Atanackovi Department of Astronomy, Faculty of Mathematics,
University of Belgrade,
Studentski trg 16, Belgrade, Serbia
Abstract. The solution of radiative transfer (RT) is a necessary
step in the complex process of an astrophysical object modeling. In
order to properly determine the physical and dynamical properties
of an object from its spectrum, a reliable numerical method that
can handle the multilevel non-LTE radiative transfer in spectral
lines is needed. This problem is difficult to solve due to the
non-local and strongly non-linear coupling between the atomic level
populations (i.e. excitation state of the gas) and the radiation
field intensities in the corresponding line transitions. For its
solution an iterative method is required. For the solution of more
complex, time-dependent or multi- dimensional RT problems,
iterative method should be as fast as possible. Here we shall
describe two methods, Iteration Factors Method (IFM) and Forth-and-
Back Implicit Lambda Iteration (FBILI) that, owing to the use of
the so called iteration factors in an explicit and implicit way
respectively, converge extremely fast to the exact solution.
Key words: Radiative transfer, Numerical methods, Stellar
atmospheres
1. INTRODUCTION
The spectra of astrophysical objects contain a wealth of
information on their chemical composition and physical properties.
The principal goal is to develop the methods to extract this
information from their spectra. In practice, the so-called
synthetic approach is applied: starting from the postulated model
of an object and the theory of radiative transfer (RT) one computes
the emergent spectrum and compares it to the observed one, produced
by the real object and its own energy transport mechanisms. The
model and the theory of radiation transport are to be modified
until the agreement between the computed and the observed spectrum
is reached. Therefore, the solution of radiative transfer in a
given astrophysical object is an inevitable step in the object
modeling.
The spectral lines are the best diagnostic tool for interpreting
the physical properties of celestial bodies. Hence, the
understanding of line formation, i.e. radiative transfer in
spectral lines is of fundamental importance and deserves a special
attention. We shall focus here on stellar spectra, since the
stellar atmospheres are the most studied and the best understood
astrophysical objects. However, the methods we consider here can be
easily generalized to the study of RT in many other media.
© Institute of Physics, 2016
Physica Macedonica, Vol. 62
At the beginning of the XX century and in the absence of advanced
computational facilities, the simplifying physical models of line
formation (e.g. pure absorption – LTE) had to be used. As the
spectral lines are formed in the upper, low density regions of
stellar atmospheres where radiative rates exceed the collisional
ones and the scattering becomes important, in 1950’s the picture of
spectrum formation was greatly improved when the assumption of
local thermodynamic equilibrium (LTE) was abandoned in favor of
statistical equilibrium or the so- called non-LTE (NLTE). The
intrinsic difficulty of the NLTE line transfer problem is in the
non- local and generally non-linear coupling between the atomic
level populations and the radiation field intensities in the
corresponding line transitions. Scattering process makes the
properties of very distant points in a medium non-locally coupled
through the radiation field. The specific intensity of the
radiation field at a given depth point depends, via the RT process,
on the physical properties over a wide range of distant points. On
the other hand, the physical state of the medium at each point
depends, via radiative excitations and ionizations, on the
radiation field itself. The dependence of the atomic level
populations on the radiation field intensity is expressed by the
statistical equilibrium (SE) or rate equations describing the
balance between all the transitions (radiative + collisional) that
populate and those that depopulate each given atomic level. From
the mathematical point of view the coupling of the radiation field
and the excitation state of the gas is described by the
simultaneous solution of the corresponding equations of radiative
transfer and statistical equilibrium.
If the line source function (the ratio of line emissivity to
opacity) can be explicitly written in terms of the mean line
radiation field intensity, as in the case of a two-level atom line
transfer, the problem is linear and can be solved by using either
direct or iterative methods. However, in the general case of a
multi-level atom line transfer, the problem is highly non-linear
and an iterative solution must be sought in order to solve the
problem.
In what follows we shall first discuss the basic equations of the
multilevel line formation problem (Sec.2), then the methods
nowadays in use for its solution (Sec.3) and, finally, two such
efficient, simple and fast convergent iterative methods that do not
require any matrix formalism – Iteration Factors Method and
Forth-and-Back Implicit Lambda Iteration (Sec.4 and 5) and,
consequently, exhibit advantages in terms of stability, as well as
considerable savings of both computational time and memory
storage.
2. THE MULTI-LEVEL ATOM LINE FORMATION PROBLEM
For the sake of an easier presentation of the methods, here we
shall consider the schematic case of a multi-level atom line
formation in a plane-parallel, stationary medium of constant
properties with complete redistribution and no background
continuum. The three-level hydrogen atom is an ideal test case for
checking the numerical accuracy and convergence properties of
various algorithms developed to solve the RT problem, and its
solution has been obtained by many authors (see e.g. [1]). Under
the above assumptions, the RT equation for any line transition ij
between two discrete energy levels i and j ( ij EE > ) takes the
form
23
O. Atanackovi: On two extremely fast methods for the solution of
the non-lte line formation proble
)]()([)( τττ τ
−= , (1)
where Iνµ (τ) is the specific intensity of the radiation field at
the mean line optical depth τ, ν is the frequency and µ is the
cosine of the angle between the direction of photon propagation and
the outward normal. The absorption line profile ν is assumed to be
independent of both τ and µ. The line source function S contains
the details of interactions between radiation and the atmospheric
gas and depends on the populations of two bound atomic levels, i
and j, the line transition is occurring between:
3
2
n A hS gnn B n B c n g
ν = =
− − . (2)
{ ( ) ( )}
( ) ( )
ij jk i ij ij k kj kj kj
n A B J C B J C n B J C n A B J C
φ φ
φ φ
Σ + + + Σ +
= Σ + + Σ + + (3)
with ji < , jk > . Here, C’s are the inelastic collisional
rates, while radiative rates contain the so-called scattering
integrals
1
1
1( ) ( ) ( ) 2
∞ ∞
−∞ −∞ −
= =∫ ∫ ∫ (4)
that account for the angle and frequency coupling of the specific
intensities at the given depth point. For the NL-level atomic
model, NL-1 linearly independent SE equations are closed by the
particle conservation equation:
i tot i
n n=∑ (5)
where totn is the total number density of atoms.
3. NUMERICAL METHODS FOR THE SOLUTION OF MULTI-LEVEL PROBLEM
The above non-linear problem requires a simultaneous solution of
the RT and SE equations, which can be achieved only through
iterations. Various methods differ in the way linearity is achieved
at each iterative step. The first robust method that could handle
mutual non-
24
Physica Macedonica, Vol. 62
linear coupling of all relevant variables in a fully
self-consistent manner was the complete linearization (CL) method
by Auer and Mihalas [2]. The coupled equations are linearized in
all variables and solved simultaneously for the variables’
corrections. The convergence rate of the CL method is very high.
However, as a very large number of frequency, angle and depth
points is usually needed for a good description of the physical
system, huge memory storage and a lot of computing time are
required. Moreover, the optimization of this global method is
rather difficult.
Much simpler and cheaper methods were needed to solve more
realistic and complex problems. These are the methods that use a
sequential iterative procedure within the so-called structural
approach (Simonneau and Crivellari [3]). They take advantage of the
fact that a problem can be divided into separate tasks according to
the physics of the problem and the mathematical properties of the
relevant equations. The current solution of each individual task is
obtained by assuming that the output of the others is known. Thus
each task can be optimized independently. More specifically, in the
multilevel line formation problem, each step of the iterative
procedure is usually split in two parts so that within each of
them, one of the two systems of equations, RT and SE, is solved by
taking the solution, or part of the solution, of the other one as
known. The coupling is performed at the end of the iterative
procedure, with the convergence rate that depends on the amount of
information transferred between the two parts.
The simplest and the well-known sequential iterative procedure is
iteration that solves the problem equations in turn. However, in
scattering-dominated media of large optical thickness (typical
non-LTE conditions), the convergence of iteration is extremely
slow. It follows the process of a particular scattering event and
at each iteration step it corrects the solutions only within a unit
optical path. It transfers from one part of the iterative step to
the other more information than necessary.
There are two broad classes of methods attempting to facilitate the
solution of NLTE line transfer problem and improving upon the
difficulties encountered by the above mentioned methods. One is
based on the so-called ETLA (Equivalent-Two-Level-Atom) approach,
described in [4], in which only one transition in the atomic model
at a time is combined with the transfer problem, whereas the
coupling of all levels is achieved by iteration over all
transitions. However, it may happen that iterative procedure
converges upon inconsistent solutions as the multilevel transition
interlocking is treated iteratively. The other approach employs an
approximate treatment of RT coupled with the full set of SE
equations. The exact operator is replaced by an approximate one,
whereas a small error introduced by this approximation is accounted
for iteratively. The approximation is based either on the physics
of the transport of photons through the medium (escape probability,
core saturation etc.) or on some computational considerations. Such
approach is known as ALI (Approximated/Accelerated Lambda
Iteration). Rybicki [5] and Cannon [6] pioneered the first ideas of
ALI methods. Rybicki eliminated the cause of the slow convergence
of simple iteration, i.e. poor conditioning of the equations
arising from a large number of scatterings, by the elimination of
scatterings in the line core (core saturation assumption) which do
not contribute much to RT. The solution of the preconditioned
equations that contain only wing components by direct iteration
converged more rapidly. Cannon was the
25
O. Atanackovi: On two extremely fast methods for the solution of
the non-lte line formation proble
first to use the idea of 'operator splitting' from the numerical
analysis in radiative transfer computations. He replaced the full
(exact) operator by the so called approximate Lambda operator - ALO
and computed a small error term made by this approximation using a
perturbation technique. His ALO was operator evaluated with fewer
angular-frequency quadrature points.
ALO has to be simpler to invert than the operator, and also as
close physical approximation of the exact operator as possible to
ensure stable and rapid convergence to the exact solution. Olson et
al. [7] were the first to point out that the diagonal (local) part
of the exact matrix represents nearly optimum approximate operator.
The most commonly used ALI method is the Jacobi iteration scheme
that employs the diagonal ALO and computes the error introduced by
this approximation iteratively. Using a 'short characteristic'
solution of the first-order differential RT equation and parabolic
approximation for the source function Olson and Kunasz [8] derived
a fast method to compute the exact diagonal of the matrix. The
approach that introduces ALO directly into SE equations to make
them linear from the outset (preconditioning) is proposed by
Rybicki and Hummer [9]. This is the well-known MALI (Multilevel
Accelerated Lambda Iteration) method, which was successfully
applied to the solution of various RT problems. The convergence
rate of the Jacobi method is usually increased by the Ng
acceleration technique [10]. The diagonal ALO is also used as the
local-operator option of the well-known multilevel transfer code
MULTI [11].
Another numerical method for NLTE RT applications that is twice as
fast as the Jacobi method is the Gauss-Seidel (GS) method by
Trujillo Bueno and Fabiani Bendicho [12]. Its implementation into
multilevel RT is made explicit by Paletou and Leger [13].
Successive Overrelaxation (SOR) is applied to speed up its
convergence. The use of additional mathematical techniques is not
straightforward. It usually needs some preliminary analyses in
order to prevent the divergence of the whole procedure. Therefore,
despite important progress made thanks to ALI methods, the search
for more efficient algorithms is still an open field of
research.
Another approach that uses the so-called iteration factors, either
in explicit or implicit way, is implemented in two extremely fast
methods: the Iteration Factors Method (IFM) and the Forth-and-Back
Implicit Lambda Iteration (FBILI), respectively. These are the
methods developed for the NLTE line transfer problem by
Atanackovic-Vukmanovic [14], Atanackovic- Vukmanovic and Simonneau
[15], Atanackovic-Vukmanovic, Crivellari and Simonneau [16] and
Kuzmanovska-Barandovska and Atanackovic [17]. They will be
discussed here in more details.
4. ITERATION FACTORS METHOD
As we have already emphasized, a global astrophysical problem can
be divided into several specific physical key-problems represented
by individual blocks of a general block- diagram. The optimum
structure of the block-diagram is determined by the physics itself.
A current solution of each block is obtained assuming that the
output from all the others is known. The solution of the global
problem is thus achieved through a sequential iterative procedure.
The most straightforward iteration scheme is very slow. However,
its slight modification can lead to an extremely high convergence
rate. One such revision is in the use of the so-called
iteration
26
Physica Macedonica, Vol. 62
factors (IFs) as the input/output of the blocks that constitute the
block diagram. At each iteration step the IFs are computed from the
current solution and then used to update it. In order to provide
fast convergence they have to be good quasi-invariants (to change
little from one iteration to another). Because being defined as the
ratios of the homologous physical quantities, they become exact
already in the first few iterations leading quickly to the exact
solution of the whole procedure.
In RT literature, the idea of iteration factors appeared for the
first time in the paper by Feautrier [18] who suggested that the
use of the ratio of two moments of the radiation field intensity
could speed up the stellar atmosphere model computations. The first
realization of the idea was the variable (depth-dependent)
Eddington factor (VEF) technique [19] developed for the solution of
the monochromatic transfer problem in plane-parallel geometry
(Fig.1). Auer and Mihalas iterated on the ratio of the third to the
first angular moments of the radiation field F=K/J, i.e. on the
so-called Eddington factor.
Figure 1. Flow-chart of the method of VEFs for the monochromatic
transfer problem.
They pointed out that it is much better to iterate on the ratio of
two quantities than on the quantities themselves, as their ratio
changes much less from one iteration to another. An extremely fast
convergence is achieved at the cost of a small additional effort,
with respect to the iteration, of iterative computation of
Eddington factor and the solution of one second-order differential
equation (i.e. two first-order differential equations closed by the
iteration factor, see Fig.1). Two to three iterations were enough
to reach the exact solution.
27
O. Atanackovi: On two extremely fast methods for the solution of
the non-lte line formation proble
The VEFs have been applied within the linearization method for
both, the stellar atmosphere modeling [2] and the line formation
[20], problems to reduce the angular dimensions of the system under
study. They enabled a complex explicit frequency-angle coupling to
be simplified, so that only the frequency coupling is treated
explicitly, whereas the angle coupling is treated iteratively.
However, a very large number of frequency points still made the
computations with the CL/VEF demanding.
The idea of VEFs is generalized in the Iteration Factors Method
(IFM), developed to solve the two-level-atom line transfer problem
by Atanackovic-Vukmanovic [14] and Atanackovic-Vukmanovic and
Simonneau [15]. For the first time they used both angle and
frequency averaged iteration factors, defined for a line as a
whole. At each iteration, depth- dependent factors defined as the
ratios of the relevant (angle and frequency) intensity moments are
computed from the current values of the radiation field and then
used to close the system of the RT equation moments. The iteration
factors method does not need any matrix operation, thus requiring
small memory storage and computational time.
The IFM is extended to the solution of the multilevel line transfer
problem by Kuzmanovska-Barandovska and Atanackovic [17]. Iterative
procedure that overcomes the above mentioned drawback of the CL/VEF
method is proposed, retaining its favorable convergence properties
while becoming computationally cheap owing to the use of iteration
factors defined for each line as a whole. The algorithm is
displayed in Fig.2.
Figure 1. Flow-chart of the IFM for the line transfer problem. The
algorithm is divided by dashed line in two parts that can be
optimized independently.
In the first part of each iteration, starting from a given source
function ( 0S ) or the one known from the previous iteration step (
iS ) we formally solve the RT equation (1) for each line
28
Physica Macedonica, Vol. 62
transition. The obtained intensities are used for the computation
of the relevant angle and frequency integrated moments and the
iteration factors as their ratios. The IFs are then used as the
known coefficients to close the system of the RT equation moments
(for all line transitions) non-linearly coupled with the
statistical equilibrium (SE) equations (for all level populations)
and to update the solution in the second part of each iteration.
Two approaches were used to solve these two systems of equations
simultaneously. The first one is based on the linearization of the
equations by neglecting the second- and higher-order terms. A
substitution of the linearized SE equations into the linearized RT
equation moments yields the system of equations that involve only
the corrections of the frequency integrated mean intensity of the
radiation field explicitly. In the second approach, the linear form
of the SE equations is obtained by assuming that the level
populations in their line-opacity-like terms are known from the
previous iteration. Thus one derives linear relations between each
line source function and the full set of the radiation field mean
intensities and solves them together with the RT equation moments
for all line transitions linearly coupled in this way. The second
approach can be labeled "preconditioning". In both approaches an
extremely fast convergence is provided and the cost per iteration
drastically reduced by the use of quasi-invariant iteration factors
in the linearly coupled RT equation moments and SE equations.
The proposed method is tested on a standard benchmark model for the
multilevel atom problems: three-level hydrogen atoms in a
plane-parallel, semi-infinite isothermal atmosphere with no
background opacity. We compared the convergence rate of the IFM
with that of the MALI method, using the results given in [9]. The
convergence rate of the IFM is higher than that of MALI by a factor
of more than 5 with respect to the solution with diagonal operator
without acceleration (D/N), and by a factor of about 2 for the
tridiagonal ALO without acceleration (T/N). Only the use of
tridiagonal ALO with Ng acceleration (T/A) provided more rapid
convergence than the IFM (that uses no additional acceleration). In
Table 1. the convergence properties of the IFM are compared with
those of the methods currently in use (MALI, Gauss-Seidel and SOR),
the results of which for the same test problem are given by Paletou
and Leger [13]. One can see that the total computational work (CW)
of the IFM is 5-6 times lesser than that of MALI. IFM is about 4
times faster than GS and 1.5 times faster than GS accelerated by
SOR.
Table 1: Comparison of the convergence properties of the IFM with
other methods (MALI, GS and SOR [13]). Iter – number of iterations
needed to satisfy a given cR (maximum relative change of the source
function between two successive iterations), CPU/Iter –
computational time per iteration and CW – total computational
work.
Rc MALI GS SOR IFM Iter CPU/Iter CW Iter CPU/Iter
CW Iter CPU/Iter
CW Iter CPU/Iter
CW 10-2 100 0.163 16.3 55 0.21 11.6 25 0.21 5.25 12 0.238
2.86
10-3 170 0.163 27.7 85 0.21 17.9 35 0.21 7.35 19 0.238 4.52
10-4 230 0.163 37.5 120 0.21 25.2 0.21 8.40 26 0.238 6.19
Very high convergence rate is not affected by the refinement of the
grid resolution, i.e. the number of iterations is practically
insensitive to the spatial grid, so that the total
computational
29
O. Atanackovi: On two extremely fast methods for the solution of
the non-lte line formation proble
work scales linearly with the number of depth grid points (see
Table 2). As an additional check the IFM was applied to the
solution of a 5-level CaII ion model, the test problem proposed by
Avrett [21], and similar, very encouraging results are obtained
[22].
5. FORTH-AND-BACK IMPLICIT LAMBDA ITERATION
Another simple and efficient method developed to accelerate the
convergence of the classical -iteration while retaining its
straightforwardness is Forth-and-Back Implicit Lambda Iteration –
FBILI [14, 16]. It makes use of the fact that even if the radiation
field is not known, its physical behavior can be easily
represented. An implicit representation of the source function is
used in the computation of both the in-going ( −
νµI ) and the out-going ( + νµI ) intensities of the
radiation field that are considered separately within a
forth-and-back approach. Namely, using the integral form of the RT
equation
dtetSeII l
/)(/ 1 )()()( (6)
and assuming polynomial (e.g. piecewise parabolic) behavior for the
source function in each depth subinterval τ , for every optical
depth point lτ one can write implicit linear relations representing
the specific intensities in terms of the yet-unknown values of the
source function and its derivatives:
)(')()( lll ScSbaI τττ νµνµνµνµ ±±±± ++= (7)
According to the idea of iteration factors, the iterative
computations of the coefficients of these implicit relations
(implicit, as the source function is a priori unknown), rather than
of the unknown functions themselves, greatly accelerate the
convergence of the iterative scheme. The coefficients are computed
layer by layer during the forward (‘in-going’) elimination process
and stored for further use in the back substitution process when
the explicit values of unknown functions are computed.
)(')() )(
Physica Macedonica, Vol. 62
where − νµa is to be computed with old (from the previous
iteration) source function. Finally, the
coefficients of the linear relation for the unknown in-going mean
radiation field
)(')()( lllll ScSbJ τττ −−− += , (9)
are stored. The aim is to obtain the coefficients of the linear
relation between the 'full' mean intensity
and the source function ( bSaJ += ) which, when substituted into
the statistical equilibrium equations, lead to the new source
function. In order to obtain the 'full' coefficients a and b , the
coefficients of the corresponding implicit relation for the
out-going mean intensities are needed. Proceeding from the lower
boundary condition we determine them simultaneously with the
updated source function when sweeping layer by layer back to the
surface. The process is iterated to convergence. A negligible
additional computational effort made within FBILI method with
respect to the classical iteration results in an extremely fast
convergence to the exact solution. This is due to the fact that the
only information transferred from the previous iteration step to
the next one is in the ratio between the non-local part of the
inward mean intensities and the current (old) source function ( 0/
Sa −
νµ ), which represents an iteration factor. The efficiency and the
accuracy of the FBILI method have been verified on several RT
problems. The two-level-atom line transfer problem with complete
redistribution (CRD) and its generalization to the case of partial
frequency redistribution (PRD), as well as the multi-level- atom
line formation problem with CRD, are considered by
Atanackovi-Vukmanovi, Crivellari and Simonneau [16]. The exact
solutions are obtained in a very small number of iterations even
under extreme non-LTE conditions. The method is generalized to the
monochromatic scattering and two-level atom line formation in
static spherical media [23, 24], to the two-level line formation in
media with low velocity fields [25] and to the two-level line
formation in 2D Cartesian coordinates [26]. The multi-level atom
line transfer in 1D is made more explicit in [27]. In Table 2, the
comparison of the convergence properties of IFM and FBILI methods
is shown. The CPU time per FBILI iteration is only 10% longer than
CPU per iteration. In terms of the total computational work, FBILI
is faster than IFM by a factor of 2. Both methods have higher
convergence rate than those currently in use (see Tables 1 and 2).
Another important advantage of the two methods with respect to the
ALI ones is that their total CW scales linearly with the number of
grid points. Table 2: Convergence properties of the IFM and FBILI
methods: the number of iterations (Iter) required to satisfy the
criterion 410−=cR , the computation time per iteration (CPU/Iter)
in seconds and total computational work (CW) for the three-level H
atom benchmark problem as a function of the number of depth points
per decade τN .
τN iteration IFM FBILI
CPU/Iter Iter CPU/Iter CW Iter CPU/Iter CW
3 0.0135 23 0.0183 0.422 11 0.0156 0.172
6 0.0264 23 0.0353 0.813 15 0.0292 0.438
9 0.0395 28 0.0525 1.469 0.0436 0.828
31
O. Atanackovi: On two extremely fast methods for the solution of
the non-lte line formation proble
6. CONCLUSIONS
In order to speed up the convergence, the iterative methods
nowadays in use to solve various RT problems usually employ some
additional mathematical acceleration techniques, the implementation
of which is not straightforward and may lead to divergence of the
iterative procedure. We have shown that the use of iteration
factors (either explicitly – Iteration Factors Method, or
implicitly – Forth-and-Back Implicit Lambda Iteration) enables an
extremely fast convergence to the exact solution of NLTE radiative
transfer problems with no need for an extra acceleration.
In particular, the advantages of the IFM with respect to the CL and
ALI methods are as follows: (1) the iteration factors defined for
line as a whole reduce drastically the need for large memory and
computing time and (2) there is no increase in the number of
iterations with the grid resolution refinement. The FBILI method is
even more simple, stable and extremely fast. The facts that it is a
two-point algorithm and that the memory storage grows linearly with
the dimensions of the problem makes FBILI very suitable for 3D
RT.
Both methods are rapidly converging, exact and stable, being thus
very promising in the applications to more complex problems, in
which RT is coupled with other physical phenomena.
ACKNOWLEDGEMENTS
This work has been realized within the project No. 176004 (”Stellar
Physics”) supported by the Ministry of Education, Science and
Technological Development of the Republic of Serbia.
REFERENCES
[1] E.H. Avrett and R. Loeser, in Numerical Radiative Transfer, ed.
W. Kalkofen (Cambridge Univ. Press), p. 135 (1987)
[2] L.H. Auer and D. Mihalas, Astrophys. J. 158, 641-655 (1969) [3]
E. Simonneau and L. Crivellari, Structural algorithms to solve
radiative transfer problems,
Research Project, Instituto de Astrofisica de Canarias (1994) [4]
D. Mihalas, Stellar Atmospheres, 2nd edition, Freeman, San
Francisco (1978) [5] G. Rybicki, in: R.G. Athay, L.L.House, G.Jr
Newkirk, editors. Line formation in the
presence of magnetic fields. High Altitude Observatory, Boulder;
p.145 (1972) [6] C.J. Cannon, Journal of Quant. Spectrosc. Radiat.
Transfer 13, 627-633 (1973) [7] G.L. Olson, L.H. Auer, J.R.
Buchler, Journal of Quant. Spectrosc. Radiat. Transfer 35,
431-
442 (1986) [8] G.L. Olson and P.B. Kunasz, Journal of Quant.
Spectrosc. Radiat. Transfer 38, 325-336
(1987) [9] G. Rybicki and D. Hummer, Astron. Astrophys. 245,
171-181 (1991) [10] K.C. Ng, J. Chem. Phys., 61, 2680-2689 (1974)
[11] M. Carlsson, A Computer Program for Solving Multi-Level
Non-LTE Radiative Transfer
Problems in Moving or Static Atmospheres, Uppsala Astron. Obs. Rep.
33, (1986, 1991) [12] J. Trujillo Bueno, P. Fabiani Bendicho,
Astrophys. J. 455, 646-657 (1995) [13] F. Paletou, L. Leger,
Journal of Quant. Spectrosc. Radiat. Transfer 103, 57-66 (2007)
[14] O. Atanackovi-Vukmanovi, PhD thesis, Department of Astronomy,
Faculty of
Mathematics, University of Belgrade (1991)
32
[15] O. Atanackovi-Vukmanovi, E. Simonneau, Journal of Quant.
Spectrosc. Radiat. Transfer 51, 525, (1994)
[16] O. Atanackovi-Vukmanovi, L. Crivellari and E. Simonneau,
Astrophys. J. 487, 735-746 (1997)
[17] O. Kuzmanovska-Barandovska and O. Atanackovi, Journal of
Quant. Spectrosc. Radiat. Transfer 111, 708-722 (2010)
[18] P.Feautrier, C.R.Acad.Sci., Paris, 258, p. 3189 (1964) [19]
L.H. Auer, D. Mihalas, MNRAS 149, 65 (1970) [20] L. Auer,
Astrophys. J. 180, 469-472 (1973) [21] E.H. Avrett, Resonance Lines
in Astrophysics, National Center for Atmospheric Research,
Boulder, 27 (1968) [22] O. Kuzmanovska-Barandovska and O.
Atanackovi, Publ. Astron. Obs. Belgrade 90, 155-
158 (2010) [23] O. Atanackovi-Vukmanovi, Serb. Astron. J. 167,
27-34 (2003) [24] O. Atanackovi-Vukmanovi, ASP Conference Series
370, 97-102 (2007) [25] I. Pirkovi and O. Atanackovi, Serb. Astron.
J. 189, 53-67 (2014) [26] I. Mili and O. Atanackovi, Advances in
Space Research 54, 1297-1307 (2014) [27] O. Kuzmanovska,
O.Atanackovi and M. Faurobert, in preparation (2016)
33
Physica Macedonica 62, (2016) p. 34-40
METHOD OF ABSOLUTE GRAVITY MEASUREMENT AND ITS INFLUENCE IN MASS
DETERMINATION
Luljeta Disha1, Defrim Bulku1 and Antoneta Deda2
1General Directorate of Metrology, Autostrada Tr-Dr, km 8,
Kashar,1000 Tirana, Albania
2 Faculty of Natural Science, Bulevardi Zogu I, 1000 Tirana,
Albania
Abstract. The determination of acceleration due to gravity has
application in many measurements. Its value is of interest in a
broad area of physical sciences, Metrology, Geophysics and Geodesy.
In Metrology, knowledge of the gravity field is very important in
mass determination, and also it is used as a transfer standard in
the determination of metrological standards involving force and
pressure measurements. The need for modern absolute gravity
measurements within 10-9 relative uncertai